On 3/13/18, 12:56 PM, "Dino Farinacci" <farina...@gmail.com> wrote:
> Using IPv6 format is something we considered while writing the draft. We
went the LCAF route to have an explicit way to (1) distinguish ILA
Identifiers/Locators from other addresses in the Mapping System, (2) specify
the Identifier/Locator length and (3) include metadata bits. However, for
simple scenarios (only ILA domain, no overlapping with non-local addresses, no
multiple SIR prefixes, fixed Identifier length, no need for metadata bits, etc)
things could work with AFI=2 format. If the rough consensus from the WG(s) is
that a plain AFI=2 format is sufficient, we can certainly update the draft. I
would like to know the opinion of others on this.
Well identifiers can be encoded as ::<64-bits> and locators can be encoded
as a regular prefix (leading bits and mask-length).
I have been running with some ILA addresses in my mapping system for a
while now. I wanted to show Tom that it could be done easily. What I did was
register a 128-bit EID which was the SIR-prefix plus identifier which mapped to
a 128-bit RLOC that contained high-order bits as the routable locator and
low-order bits as the identifier. I realize this is a bit redundant, but it
could be done with no protocol or implementation changes.
There is another option that we discussed when we were considering IPv6
encoding. Instead of being redundant with the low-order bits of the Locator,
you could use those to encode a "special identifier" that you will use when
sending control-plane messages to that Locator. This way, the ILA device at
that Locator has a clear way to punt packets to control-plane processing.
lisp mailing list